ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/libecb/ecb.pod
Revision: 1.42
Committed: Mon May 28 08:54:03 2012 UTC (12 years, 1 month ago) by root
Branch: MAIN
Changes since 1.41: +3 -3 lines
Log Message:
*** empty log message ***

File Contents

# User Rev Content
1 root 1.14 =head1 LIBECB - e-C-Builtins
2 root 1.3
3 root 1.14 =head2 ABOUT LIBECB
4    
5     Libecb is currently a simple header file that doesn't require any
6     configuration to use or include in your project.
7    
8 sf-exg 1.16 It's part of the e-suite of libraries, other members of which include
9 root 1.14 libev and libeio.
10    
11     Its homepage can be found here:
12    
13     http://software.schmorp.de/pkg/libecb
14    
15     It mainly provides a number of wrappers around GCC built-ins, together
16     with replacement functions for other compilers. In addition to this,
17 sf-exg 1.16 it provides a number of other lowlevel C utilities, such as endianness
18 root 1.14 detection, byte swapping or bit rotations.
19    
20 root 1.24 Or in other words, things that should be built into any standard C system,
21     but aren't, implemented as efficient as possible with GCC, and still
22     correct with other compilers.
23 root 1.17
24 root 1.14 More might come.
25 root 1.3
26     =head2 ABOUT THE HEADER
27    
28 root 1.14 At the moment, all you have to do is copy F<ecb.h> somewhere where your
29     compiler can find it and include it:
30    
31     #include <ecb.h>
32    
33     The header should work fine for both C and C++ compilation, and gives you
34     all of F<inttypes.h> in addition to the ECB symbols.
35    
36 sf-exg 1.16 There are currently no object files to link to - future versions might
37 root 1.14 come with an (optional) object code library to link against, to reduce
38     code size or gain access to additional features.
39    
40     It also currently includes everything from F<inttypes.h>.
41    
42     =head2 ABOUT THIS MANUAL / CONVENTIONS
43    
44     This manual mainly describes each (public) function available after
45     including the F<ecb.h> header. The header might define other symbols than
46     these, but these are not part of the public API, and not supported in any
47     way.
48    
49     When the manual mentions a "function" then this could be defined either as
50     as inline function, a macro, or an external symbol.
51    
52     When functions use a concrete standard type, such as C<int> or
53     C<uint32_t>, then the corresponding function works only with that type. If
54     only a generic name is used (C<expr>, C<cond>, C<value> and so on), then
55     the corresponding function relies on C to implement the correct types, and
56     is usually implemented as a macro. Specifically, a "bool" in this manual
57     refers to any kind of boolean value, not a specific type.
58 root 1.1
59 root 1.40 =head2 TYPES / TYPE SUPPORT
60    
61     ecb.h makes sure that the following types are defined (in the expected way):
62    
63 root 1.42 int8_t uint8_t int16_t uint16_t
64     int32_t uint32_t int64_t uint64_t
65     intptr_t uintptr_t ptrdiff_t
66 root 1.40
67     The macro C<ECB_PTRSIZE> is defined to the size of a pointer on this
68     platform (currently C<4> or C<8>).
69    
70 root 1.1 =head2 GCC ATTRIBUTES
71    
72 root 1.20 A major part of libecb deals with GCC attributes. These are additional
73 sf-exg 1.26 attributes that you can assign to functions, variables and sometimes even
74 root 1.20 types - much like C<const> or C<volatile> in C.
75    
76     While GCC allows declarations to show up in many surprising places,
77 sf-exg 1.26 but not in many expected places, the safest way is to put attribute
78 root 1.20 declarations before the whole declaration:
79    
80     ecb_const int mysqrt (int a);
81     ecb_unused int i;
82    
83     For variables, it is often nicer to put the attribute after the name, and
84     avoid multiple declarations using commas:
85    
86     int i ecb_unused;
87 root 1.3
88 root 1.1 =over 4
89    
90 root 1.2 =item ecb_attribute ((attrs...))
91 root 1.1
92 root 1.15 A simple wrapper that expands to C<__attribute__((attrs))> on GCC, and to
93     nothing on other compilers, so the effect is that only GCC sees these.
94    
95     Example: use the C<deprecated> attribute on a function.
96    
97     ecb_attribute((__deprecated__)) void
98     do_not_use_me_anymore (void);
99 root 1.2
100 root 1.3 =item ecb_unused
101    
102     Marks a function or a variable as "unused", which simply suppresses a
103     warning by GCC when it detects it as unused. This is useful when you e.g.
104     declare a variable but do not always use it:
105    
106 root 1.15 {
107     int var ecb_unused;
108 root 1.3
109 root 1.15 #ifdef SOMECONDITION
110     var = ...;
111     return var;
112     #else
113     return 0;
114     #endif
115     }
116 root 1.3
117 root 1.31 =item ecb_inline
118 root 1.29
119     This is not actually an attribute, but you use it like one. It expands
120     either to C<static inline> or to just C<static>, if inline isn't
121     supported. It should be used to declare functions that should be inlined,
122     for code size or speed reasons.
123    
124     Example: inline this function, it surely will reduce codesize.
125    
126 root 1.31 ecb_inline int
127 root 1.29 negmul (int a, int b)
128     {
129     return - (a * b);
130     }
131    
132 root 1.2 =item ecb_noinline
133    
134 root 1.9 Prevent a function from being inlined - it might be optimised away, but
135 root 1.3 not inlined into other functions. This is useful if you know your function
136     is rarely called and large enough for inlining not to be helpful.
137    
138 root 1.2 =item ecb_noreturn
139    
140 root 1.17 Marks a function as "not returning, ever". Some typical functions that
141     don't return are C<exit> or C<abort> (which really works hard to not
142     return), and now you can make your own:
143    
144     ecb_noreturn void
145     my_abort (const char *errline)
146     {
147     puts (errline);
148     abort ();
149     }
150    
151 sf-exg 1.19 In this case, the compiler would probably be smart enough to deduce it on
152     its own, so this is mainly useful for declarations.
153 root 1.17
154 root 1.2 =item ecb_const
155    
156 sf-exg 1.19 Declares that the function only depends on the values of its arguments,
157 root 1.17 much like a mathematical function. It specifically does not read or write
158     any memory any arguments might point to, global variables, or call any
159     non-const functions. It also must not have any side effects.
160    
161     Such a function can be optimised much more aggressively by the compiler -
162     for example, multiple calls with the same arguments can be optimised into
163     a single call, which wouldn't be possible if the compiler would have to
164     expect any side effects.
165    
166     It is best suited for functions in the sense of mathematical functions,
167 sf-exg 1.19 such as a function returning the square root of its input argument.
168 root 1.17
169     Not suited would be a function that calculates the hash of some memory
170     area you pass in, prints some messages or looks at a global variable to
171     decide on rounding.
172    
173     See C<ecb_pure> for a slightly less restrictive class of functions.
174    
175 root 1.2 =item ecb_pure
176    
177 root 1.17 Similar to C<ecb_const>, declares a function that has no side
178     effects. Unlike C<ecb_const>, the function is allowed to examine global
179     variables and any other memory areas (such as the ones passed to it via
180     pointers).
181    
182     While these functions cannot be optimised as aggressively as C<ecb_const>
183     functions, they can still be optimised away in many occasions, and the
184     compiler has more freedom in moving calls to them around.
185    
186     Typical examples for such functions would be C<strlen> or C<memcmp>. A
187     function that calculates the MD5 sum of some input and updates some MD5
188     state passed as argument would I<NOT> be pure, however, as it would modify
189     some memory area that is not the return value.
190    
191 root 1.2 =item ecb_hot
192    
193 root 1.17 This declares a function as "hot" with regards to the cache - the function
194     is used so often, that it is very beneficial to keep it in the cache if
195     possible.
196    
197     The compiler reacts by trying to place hot functions near to each other in
198     memory.
199    
200 sf-exg 1.19 Whether a function is hot or not often depends on the whole program,
201 root 1.17 and less on the function itself. C<ecb_cold> is likely more useful in
202     practise.
203    
204 root 1.2 =item ecb_cold
205    
206 root 1.17 The opposite of C<ecb_hot> - declares a function as "cold" with regards to
207     the cache, or in other words, this function is not called often, or not at
208     speed-critical times, and keeping it in the cache might be a waste of said
209     cache.
210    
211     In addition to placing cold functions together (or at least away from hot
212     functions), this knowledge can be used in other ways, for example, the
213     function will be optimised for size, as opposed to speed, and codepaths
214     leading to calls to those functions can automatically be marked as if
215 root 1.27 C<ecb_expect_false> had been used to reach them.
216 root 1.17
217     Good examples for such functions would be error reporting functions, or
218     functions only called in exceptional or rare cases.
219    
220 root 1.2 =item ecb_artificial
221    
222 root 1.17 Declares the function as "artificial", in this case meaning that this
223     function is not really mean to be a function, but more like an accessor
224     - many methods in C++ classes are mere accessor functions, and having a
225     crash reported in such a method, or single-stepping through them, is not
226     usually so helpful, especially when it's inlined to just a few instructions.
227    
228     Marking them as artificial will instruct the debugger about just this,
229     leading to happier debugging and thus happier lives.
230    
231     Example: in some kind of smart-pointer class, mark the pointer accessor as
232     artificial, so that the whole class acts more like a pointer and less like
233     some C++ abstraction monster.
234    
235     template<typename T>
236     struct my_smart_ptr
237     {
238     T *value;
239    
240     ecb_artificial
241     operator T *()
242     {
243     return value;
244     }
245     };
246    
247 root 1.2 =back
248 root 1.1
249     =head2 OPTIMISATION HINTS
250    
251     =over 4
252    
253 root 1.14 =item bool ecb_is_constant(expr)
254 root 1.1
255 root 1.3 Returns true iff the expression can be deduced to be a compile-time
256     constant, and false otherwise.
257    
258     For example, when you have a C<rndm16> function that returns a 16 bit
259     random number, and you have a function that maps this to a range from
260 root 1.5 0..n-1, then you could use this inline function in a header file:
261 root 1.3
262     ecb_inline uint32_t
263     rndm (uint32_t n)
264     {
265 root 1.6 return (n * (uint32_t)rndm16 ()) >> 16;
266 root 1.3 }
267    
268     However, for powers of two, you could use a normal mask, but that is only
269     worth it if, at compile time, you can detect this case. This is the case
270     when the passed number is a constant and also a power of two (C<n & (n -
271     1) == 0>):
272    
273     ecb_inline uint32_t
274     rndm (uint32_t n)
275     {
276     return is_constant (n) && !(n & (n - 1))
277     ? rndm16 () & (num - 1)
278 root 1.6 : (n * (uint32_t)rndm16 ()) >> 16;
279 root 1.3 }
280    
281 root 1.14 =item bool ecb_expect (expr, value)
282 root 1.1
283 root 1.7 Evaluates C<expr> and returns it. In addition, it tells the compiler that
284     the C<expr> evaluates to C<value> a lot, which can be used for static
285     branch optimisations.
286 root 1.1
287 root 1.27 Usually, you want to use the more intuitive C<ecb_expect_true> and
288     C<ecb_expect_false> functions instead.
289 root 1.1
290 root 1.27 =item bool ecb_expect_true (cond)
291 root 1.1
292 root 1.27 =item bool ecb_expect_false (cond)
293 root 1.1
294 root 1.7 These two functions expect a expression that is true or false and return
295     C<1> or C<0>, respectively, so when used in the condition of an C<if> or
296     other conditional statement, it will not change the program:
297    
298     /* these two do the same thing */
299     if (some_condition) ...;
300 root 1.27 if (ecb_expect_true (some_condition)) ...;
301 root 1.7
302 root 1.27 However, by using C<ecb_expect_true>, you tell the compiler that the
303     condition is likely to be true (and for C<ecb_expect_false>, that it is
304     unlikely to be true).
305 root 1.7
306 root 1.9 For example, when you check for a null pointer and expect this to be a
307 root 1.27 rare, exceptional, case, then use C<ecb_expect_false>:
308 root 1.7
309     void my_free (void *ptr)
310     {
311 root 1.27 if (ecb_expect_false (ptr == 0))
312 root 1.7 return;
313     }
314    
315     Consequent use of these functions to mark away exceptional cases or to
316     tell the compiler what the hot path through a function is can increase
317     performance considerably.
318    
319 root 1.27 You might know these functions under the name C<likely> and C<unlikely>
320     - while these are common aliases, we find that the expect name is easier
321     to understand when quickly skimming code. If you wish, you can use
322     C<ecb_likely> instead of C<ecb_expect_true> and C<ecb_unlikely> instead of
323     C<ecb_expect_false> - these are simply aliases.
324    
325 root 1.7 A very good example is in a function that reserves more space for some
326     memory block (for example, inside an implementation of a string stream) -
327 root 1.9 each time something is added, you have to check for a buffer overrun, but
328 root 1.7 you expect that most checks will turn out to be false:
329    
330     /* make sure we have "size" extra room in our buffer */
331     ecb_inline void
332     reserve (int size)
333     {
334 root 1.27 if (ecb_expect_false (current + size > end))
335 root 1.7 real_reserve_method (size); /* presumably noinline */
336     }
337    
338 root 1.14 =item bool ecb_assume (cond)
339 root 1.7
340     Try to tell the compiler that some condition is true, even if it's not
341     obvious.
342    
343     This can be used to teach the compiler about invariants or other
344     conditions that might improve code generation, but which are impossible to
345     deduce form the code itself.
346    
347 root 1.27 For example, the example reservation function from the C<ecb_expect_false>
348 root 1.7 description could be written thus (only C<ecb_assume> was added):
349    
350     ecb_inline void
351     reserve (int size)
352     {
353 root 1.27 if (ecb_expect_false (current + size > end))
354 root 1.7 real_reserve_method (size); /* presumably noinline */
355    
356     ecb_assume (current + size <= end);
357     }
358    
359     If you then call this function twice, like this:
360    
361     reserve (10);
362     reserve (1);
363    
364     Then the compiler I<might> be able to optimise out the second call
365     completely, as it knows that C<< current + 1 > end >> is false and the
366     call will never be executed.
367    
368     =item bool ecb_unreachable ()
369    
370     This function does nothing itself, except tell the compiler that it will
371 root 1.9 never be executed. Apart from suppressing a warning in some cases, this
372 root 1.7 function can be used to implement C<ecb_assume> or similar functions.
373    
374 root 1.14 =item bool ecb_prefetch (addr, rw, locality)
375 root 1.7
376     Tells the compiler to try to prefetch memory at the given C<addr>ess
377 root 1.10 for either reading (C<rw> = 0) or writing (C<rw> = 1). A C<locality> of
378 root 1.7 C<0> means that there will only be one access later, C<3> means that
379     the data will likely be accessed very often, and values in between mean
380     something... in between. The memory pointed to by the address does not
381     need to be accessible (it could be a null pointer for example), but C<rw>
382     and C<locality> must be compile-time constants.
383    
384     An obvious way to use this is to prefetch some data far away, in a big
385 root 1.9 array you loop over. This prefetches memory some 128 array elements later,
386 root 1.7 in the hope that it will be ready when the CPU arrives at that location.
387    
388     int sum = 0;
389    
390     for (i = 0; i < N; ++i)
391     {
392     sum += arr [i]
393     ecb_prefetch (arr + i + 128, 0, 0);
394     }
395    
396     It's hard to predict how far to prefetch, and most CPUs that can prefetch
397     are often good enough to predict this kind of behaviour themselves. It
398     gets more interesting with linked lists, especially when you do some fair
399     processing on each list element:
400    
401     for (node *n = start; n; n = n->next)
402     {
403     ecb_prefetch (n->next, 0, 0);
404     ... do medium amount of work with *n
405     }
406    
407     After processing the node, (part of) the next node might already be in
408     cache.
409 root 1.1
410 root 1.2 =back
411 root 1.1
412 root 1.36 =head2 BIT FIDDLING / BIT WIZARDRY
413 root 1.1
414 root 1.4 =over 4
415    
416 root 1.3 =item bool ecb_big_endian ()
417    
418     =item bool ecb_little_endian ()
419    
420 sf-exg 1.11 These two functions return true if the byte order is big endian
421     (most-significant byte first) or little endian (least-significant byte
422     first) respectively.
423    
424 root 1.24 On systems that are neither, their return values are unspecified.
425    
426 root 1.3 =item int ecb_ctz32 (uint32_t x)
427    
428 root 1.35 =item int ecb_ctz64 (uint64_t x)
429    
430 sf-exg 1.11 Returns the index of the least significant bit set in C<x> (or
431 root 1.24 equivalently the number of bits set to 0 before the least significant bit
432 root 1.35 set), starting from 0. If C<x> is 0 the result is undefined.
433    
434 root 1.36 For smaller types than C<uint32_t> you can safely use C<ecb_ctz32>.
435    
436 root 1.35 For example:
437 sf-exg 1.11
438 root 1.15 ecb_ctz32 (3) = 0
439     ecb_ctz32 (6) = 1
440 sf-exg 1.11
441 root 1.41 =item bool ecb_is_pot32 (uint32_t x)
442    
443     =item bool ecb_is_pot64 (uint32_t x)
444    
445     Return true iff C<x> is a power of two or C<x == 0>.
446    
447     For smaller types then C<uint32_t> you can safely use C<ecb_is_pot32>.
448    
449 root 1.35 =item int ecb_ld32 (uint32_t x)
450    
451     =item int ecb_ld64 (uint64_t x)
452    
453     Returns the index of the most significant bit set in C<x>, or the number
454     of digits the number requires in binary (so that C<< 2**ld <= x <
455     2**(ld+1) >>). If C<x> is 0 the result is undefined. A common use case is
456     to compute the integer binary logarithm, i.e. C<floor (log2 (n))>, for
457     example to see how many bits a certain number requires to be encoded.
458    
459     This function is similar to the "count leading zero bits" function, except
460     that that one returns how many zero bits are "in front" of the number (in
461     the given data type), while C<ecb_ld> returns how many bits the number
462     itself requires.
463    
464 root 1.36 For smaller types than C<uint32_t> you can safely use C<ecb_ld32>.
465    
466 root 1.3 =item int ecb_popcount32 (uint32_t x)
467    
468 root 1.35 =item int ecb_popcount64 (uint64_t x)
469    
470 root 1.36 Returns the number of bits set to 1 in C<x>.
471    
472     For smaller types than C<uint32_t> you can safely use C<ecb_popcount32>.
473    
474     For example:
475 sf-exg 1.11
476 root 1.15 ecb_popcount32 (7) = 3
477     ecb_popcount32 (255) = 8
478 sf-exg 1.11
479 root 1.39 =item uint8_t ecb_bitrev8 (uint8_t x)
480    
481     =item uint16_t ecb_bitrev16 (uint16_t x)
482    
483     =item uint32_t ecb_bitrev32 (uint32_t x)
484    
485     Reverses the bits in x, i.e. the MSB becomes the LSB, MSB-1 becomes LSB+1
486     and so on.
487    
488     Example:
489    
490     ecb_bitrev8 (0xa7) = 0xea
491     ecb_bitrev32 (0xffcc4411) = 0x882233ff
492    
493 root 1.8 =item uint32_t ecb_bswap16 (uint32_t x)
494    
495 root 1.3 =item uint32_t ecb_bswap32 (uint32_t x)
496    
497 root 1.34 =item uint64_t ecb_bswap64 (uint64_t x)
498 sf-exg 1.13
499 root 1.34 These functions return the value of the 16-bit (32-bit, 64-bit) value
500     C<x> after reversing the order of bytes (0x11223344 becomes 0x44332211 in
501     C<ecb_bswap32>).
502    
503     =item uint8_t ecb_rotl8 (uint8_t x, unsigned int count)
504    
505     =item uint16_t ecb_rotl16 (uint16_t x, unsigned int count)
506 root 1.3
507     =item uint32_t ecb_rotl32 (uint32_t x, unsigned int count)
508    
509 root 1.34 =item uint64_t ecb_rotl64 (uint64_t x, unsigned int count)
510    
511     =item uint8_t ecb_rotr8 (uint8_t x, unsigned int count)
512    
513     =item uint16_t ecb_rotr16 (uint16_t x, unsigned int count)
514    
515     =item uint32_t ecb_rotr32 (uint32_t x, unsigned int count)
516    
517 root 1.33 =item uint64_t ecb_rotr64 (uint64_t x, unsigned int count)
518    
519 root 1.34 These two families of functions return the value of C<x> after rotating
520     all the bits by C<count> positions to the right (C<ecb_rotr>) or left
521     (C<ecb_rotl>).
522 sf-exg 1.11
523 root 1.20 Current GCC versions understand these functions and usually compile them
524 root 1.34 to "optimal" code (e.g. a single C<rol> or a combination of C<shld> on
525     x86).
526 root 1.20
527 root 1.3 =back
528 root 1.1
529     =head2 ARITHMETIC
530    
531 root 1.3 =over 4
532    
533 root 1.14 =item x = ecb_mod (m, n)
534 root 1.3
535 root 1.25 Returns C<m> modulo C<n>, which is the same as the positive remainder
536     of the division operation between C<m> and C<n>, using floored
537     division. Unlike the C remainder operator C<%>, this function ensures that
538     the return value is always positive and that the two numbers I<m> and
539     I<m' = m + i * n> result in the same value modulo I<n> - in other words,
540     C<ecb_mod> implements the mathematical modulo operation, which is missing
541     in the language.
542 root 1.14
543 sf-exg 1.23 C<n> must be strictly positive (i.e. C<< >= 1 >>), while C<m> must be
544 root 1.14 negatable, that is, both C<m> and C<-m> must be representable in its
545 root 1.30 type (this typically excludes the minimum signed integer value, the same
546 root 1.25 limitation as for C</> and C<%> in C).
547 sf-exg 1.11
548 root 1.24 Current GCC versions compile this into an efficient branchless sequence on
549 root 1.28 almost all CPUs.
550 root 1.24
551     For example, when you want to rotate forward through the members of an
552     array for increasing C<m> (which might be negative), then you should use
553     C<ecb_mod>, as the C<%> operator might give either negative results, or
554     change direction for negative values:
555    
556     for (m = -100; m <= 100; ++m)
557     int elem = myarray [ecb_mod (m, ecb_array_length (myarray))];
558    
559 sf-exg 1.37 =item x = ecb_div_rd (val, div)
560    
561     =item x = ecb_div_ru (val, div)
562    
563     Returns C<val> divided by C<div> rounded down or up, respectively.
564     C<val> and C<div> must have integer types and C<div> must be strictly
565 sf-exg 1.38 positive. Note that these functions are implemented with macros in C
566     and with function templates in C++.
567 sf-exg 1.37
568 root 1.3 =back
569 root 1.1
570     =head2 UTILITY
571    
572 root 1.3 =over 4
573    
574 sf-exg 1.23 =item element_count = ecb_array_length (name)
575 root 1.3
576 sf-exg 1.13 Returns the number of elements in the array C<name>. For example:
577    
578     int primes[] = { 2, 3, 5, 7, 11 };
579     int sum = 0;
580    
581     for (i = 0; i < ecb_array_length (primes); i++)
582     sum += primes [i];
583    
584 root 1.3 =back
585 root 1.1
586